Hi Guenter! Yes, I can confirm that the bits are always zero, even if the bank is set to an invalid number. I used i2cset/i2cdump and regardless of the bank selection (I tried 5, 6, 0x32, 0x66 and 0xff) the higher bits are read back as zero. However, the lower bits return the selected bank number even if it is invalid (i.e. 5, 6, 2, 6 and 7 for the above tests). The ID bytes (0x7a-0x7c) are always correct (0x50, 0xc5, 0x51) for invalid banks as well (the rest of the bytes are 0xff's). Thanks. Vadim. On Fri, 2015-02-27 at 08:36 -0800, Guenter Roeck wrote: > The bank register has five unused bits. Verify that those bits are zero > to strenghten the detect function. > > Cc: Vadim V. Vlasov <vvlasov@xxxxxxxxxxxxx> > Signed-off-by: Guenter Roeck <linux@xxxxxxxxxxxx> > --- > Vadim, it would be great if you can confirm that/if this works. > If yes, I'll use the same code in sensors-detect. > > Thanks, > Guenter > > drivers/hwmon/nct7904.c | 3 ++- > 1 file changed, 2 insertions(+), 1 deletion(-) > > diff --git a/drivers/hwmon/nct7904.c b/drivers/hwmon/nct7904.c > index eaa8234..b77b82f 100644 > --- a/drivers/hwmon/nct7904.c > +++ b/drivers/hwmon/nct7904.c > @@ -502,7 +502,8 @@ static int nct7904_detect(struct i2c_client *client, > /* Determine the chip type. */ > if (i2c_smbus_read_byte_data(client, VENDOR_ID_REG) != NUVOTON_ID || > i2c_smbus_read_byte_data(client, CHIP_ID_REG) != NCT7904_ID || > - (i2c_smbus_read_byte_data(client, DEVICE_ID_REG) & 0xf0) != 0x50) > + (i2c_smbus_read_byte_data(client, DEVICE_ID_REG) & 0xf0) != 0x50 || > + (i2c_smbus_read_byte_data(client, BANK_SEL_REG) & 0xf8) != 0x00) > return -ENODEV; > > strlcpy(info->type, "nct7904", I2C_NAME_SIZE); _______________________________________________ lm-sensors mailing list lm-sensors@xxxxxxxxxxxxxx http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
